day-trader-skill README

项目的 README,来自 github.com/Jakeschincariol/day-trader-skill。合并为一个 markdown 文件:README.md。

The Day Trader skill

11 Claude skills for day trading: scan thousands of charts, track breaking news, time every entry and exit, and replay every outcome before you risk a dollar.

Free, MIT, no signup, no API key. The tools are plain Python that read free public data: Yahoo Finance charts, Nasdaq Trader's symbol lists, SEC EDGAR and the House Clerk.

One scans thousands of charts for setups. Another tracks breaking news on every stock you watch. One watches a setup bar by bar and tells you the moment to get in, and another sets your stop and your target before you do. And the craziest one is the Time Machine: before you risk a dollar, it replays the rest of the session 10,000 times from the stock's own recent bars, plus every past day the same setup fired, and tells you the honest odds.

And no trade goes through until you say yes. Nothing in this pack can place an order on its own. Paper trading is the default.

Looking for the original Trader skill? Copy an investor's 13F, the stock screener, tax lots and rebalancing live in github.com/Jakeschincariol/trader-skill. This repo is its day-trading sequel: the Trader skill v2.

Install

Paste this into Claude:

https://github.com/Jakeschincariol/day-trader-skill

install skill

Or as a plugin, in Claude Code:

/plugin marketplace add Jakeschincariol/day-trader-skill
/plugin install day-trader-skill@day-trader-skill

Claude Code namespaces plugin skills, so installed as a plugin they show up as /day-trader-skill:trader-scanner and so on. Copy the folders instead if you want plain /trader-scanner:

git clone https://github.com/Jakeschincariol/day-trader-skill.git
cp -r day-trader-skill/trader-* ~/.claude/skills/

Project-local instead of global: copy the same folders into your repo's .claude/skills/. No Claude Code at all? Paste any single SKILL.md at the top of a chat and it runs as a mode. You lose the Python tools, but the method works.

The tools need Python 3.8 or newer. Nothing to pip install. The SEC tools (news filings, the analyst) ask you to identify yourself: put your name and email in trader/config.json ("sec_user_agent": "Your Name [email protected]") or in the SEC_USER_AGENT environment variable.

The eleven

command job what it does
/trader-scanner Chart Scanner Scans thousands of charts for setups: opening-range breakouts, flags, oversold bounces.
/trader-news News Monitor Tracks breaking news on every stock you watch and flags what moves it.
/trader-entry Entry Timer Watches a setup bar by bar and tells you the moment to get in.
/trader-exit Exit Planner Sets your stop and your target before you enter, so you know when to get out.
/trader-timemachine Time Machine Replays every possible outcome of a trade before you risk a dollar: 10,000 simulated sessions plus every past time the setup fired.
/trader-risk Risk Manager Checks every trade against your limits and stops anything over them.
/trader-budget Budget Manager Sizes every position to your budget and your max loss per trade.
/trader-broker Order Executor Sends approved orders to your broker, paper trading by default, never without your yes.
/trader-alerts Price Alerts Rings when a price, a stop or the opening bell hits.
/trader-analyst Earnings Analyst Reads earnings reports and sums up what changed.
/trader-pelosi Pelosi Tracker Reads Congress's stock-trade disclosures the minute they are filed.

How to use it

Everything lands in a trader/ folder in your project, and each skill reads what the last one wrote:

scanner -> news -> entry -> exit -> timemachine -> budget -> risk -> broker   (one trade)
alerts: stop, target, flat by the close                                       (while it is open)
news / analyst / pelosi                                                        (on a schedule)

An example, the trade in the reel: Tesla, Monday October 19 2026, real data.

  1. /trader-scanner at 10:15 AM ET: of the 231 charts in the reel's scan, 42 had closed above their 15-minute opening range. TSLA is one: range 354.47 to 362.18, the 10:10 candle closed at 364.18, 0.55 % over the top.
  2. /trader-news: the morning's TSLA headlines, in New York time ("Tesla-SpaceX Merger Mania Strikes Wall Street Ahead of Earnings", 8:04, tagged earnings and deal). Tesla reports this week: a day trade is out before the close, so it is not exposed to the report.
  3. /trader-entry watched TSLA bar by bar. At 10:12 it said waiting (the 10:10 bar was still forming). At 10:15 it said triggered: the first 5-minute close above the range, 364.18.
  4. /trader-exit set the plan before the entry: stop at the range's midpoint, 358.32, so 5.86 a share at risk; target one times that risk, 370.04; flat by 3:55 PM either way.
  5. /trader-timemachine replayed the rest of the day 10,000 times, each 5-minute candle drawn from TSLA's own candles at that time of day over the previous 59 sessions (seed 19, so anyone can repeat it): target first 29.4 %, stop first 25.5 %, still open at the close 45.1 %. It also re-ran the same rule on those 59 real sessions: it fired 33 times, 13 hit the target, 15 the stop, 5 neither. Its verdict: close to a coin flip. That is the point of looking before you risk a dollar.
  6. /trader-budget with a $5,000 account and a $50 max loss per trade: 50 / 5.86 = 8 shares ($2,913.44). If the stop fills at 358.32, the loss is $46.88. It writes one bracket order: buy 8 TSLA, stop-loss 358.32, take-profit 370.04.
  7. /trader-risk checks it against your limits: the max loss, the daily loss limit, trades per day, reward to risk, the entry window, position size and cash. PASS or BLOCK, with the reason.
  8. /trader-broker shows you the ticket in plain words. You type yes for that exact order. Claude calls your brokerage connector (Alpaca's MCP server, paper account by default), and Claude's own "Allow once / Deny" card asks you one more time.
  9. /trader-alerts rings on the stop, the target and at 3:55 PM to flatten whatever is still open.

The tools

All standard-library Python. The data tools read public endpoints and are polite about it.

python3 trader-scanner/scan.py universe                                # ~5,900 US stocks, Nasdaq Trader
python3 trader-scanner/scan.py daily                                   # pre-market setups, whole universe
python3 trader-scanner/scan.py orb --at 10:15                          # opening-range breakouts so far
python3 trader-news/news.py --tickers TSLA --hours 12                  # headlines, 8-Ks, today's move
python3 trader-entry/entry.py check TSLA --setup orb                   # waiting / triggered / expired
python3 trader-entry/entry.py watch TSLA --every 30                    # bar by bar until it triggers
python3 trader-exit/exit.py plan --stop or-mid --target 1R --write     # -> trader/plan.json
python3 trader-timemachine/timemachine.py run --write                  # 10,000 replays + the history
python3 trader-budget/size.py --write                                  # -> trader/orders.json
python3 trader-risk/check.py                                           # PASS / BLOCK per order
python3 trader-broker/ticket.py show                                   # the tickets, READY or not
python3 trader-alerts/alerts.py add --from-plan                        # stop, target, flat by close
python3 trader-alerts/alerts.py check                                  # rings what hit, once
python3 trader-analyst/facts.py TSLA                                   # quarters, growth, EPS
python3 trader-pelosi/ptr.py index --member pelosi                     # her PTRs this year
python3 -m unittest discover -s tests -v

The tests run offline, on fixtures cut from real data: TSLA's 5-minute bars for the 60 days to October 19 2026, the 10:15 AM opening-range scan, Yahoo's TSLA headlines that morning, the House Clerk's 2026 index, real PTR text and real XBRL facts. They reproduce every number in the example above.

Fine print

Not financial advice. This is software that reads public data and does arithmetic. It does not know your situation and it cannot see the future. Most people who day trade lose money. You can lose money. Decide for yourself, and talk to a licensed professional if you need advice.

No trade goes through until you say yes. Only /trader-broker can lead to an order, and only after /trader-risk passes that exact order and you type yes for it in the chat. Set your brokerage tool to "Needs approval" in Claude (Customize, Connectors, tool permissions), never "Always allow", so Claude's own "Allow once / Deny" card is a second check. No options, margin, short selling or leverage unless you turn them on.

Paper first. "paper": true is the default in config.example.json, and Alpaca's MCP server defaults to its paper account. Keys live in environment variables, never in the chat.

The Time Machine is a simulation, not a prediction. It is built from the stock's own last 60 days of 5-minute bars. The future can do things those 60 days never did: an earnings report, a halt, a headline. Treat its odds as a reality check, never as a promise.

A stop is a plan, not a guarantee. In a fast market or across a gap, a stop order can fill well past its price. The max loss shown is the loss at the stop price.

The data is free and unofficial. Yahoo Finance's chart endpoint is not an official API: it can lag, change or rate-limit you. The scanner and the entry timer read 1-minute or 5-minute bars, not every tick. Check anything that matters against your broker before you act on it.

Day-trading rules. FINRA removed the $25,000 pattern-day-trader minimum on June 4, 2026 (Regulatory Notice 26-10), and brokers may phase the change in until October 20, 2027, so yours may still apply the old rule. If it does, set "pdt_guard": true and /trader-risk counts your day trades.

Congress's trades are public, not inside information. Members must disclose trades over $1,000 within 45 days (the STOCK Act). The Pelosi skill reads those public filings the minute they post.

Taxes. Day-trading gains are short-term gains, and frequent trading makes wash sales likely. Nothing here is tax advice.

Files

trader-scanner/     SKILL.md, scan.py
trader-news/        SKILL.md, news.py
trader-entry/       SKILL.md, entry.py
trader-exit/        SKILL.md, exit.py
trader-timemachine/ SKILL.md, timemachine.py
trader-risk/        SKILL.md, check.py
trader-budget/      SKILL.md, size.py, config.example.json, targets.example.json
trader-broker/      SKILL.md, ticket.py, claude_desktop_config.example.json
trader-alerts/      SKILL.md, alerts.py, alerts.example.json
trader-analyst/     SKILL.md, facts.py, metrics.example.json
trader-pelosi/      SKILL.md, ptr.py
tests/              unittest suite + offline fixtures

MIT. Made by Jake Schincariol.

day-trader-skill (Jakeschincariol/day-trader-skill)

正在搜索数据库…

这是什么?一套包含 11 个 Claude 技能的工具包,用于日内交易。它可以扫描图表寻找交易机会,追踪突发新闻,帮你判断入场和出场时机,并让你在动用真金白银之前,把一笔交易回放数千次。
这是一个交易应用,还是在 Claude 里运行的东西?它是一组 Claude 技能(包含 SKILL.md 文件和简单的 Python 辅助脚本),在 Claude 或 Claude Code 中运行,并不是独立的应用程序。Python 工具只是读取 Yahoo Finance、SEC EDGAR 和 House Clerk 等免费公开的数据源。
使用它需要券商账户或 API 密钥吗?市场数据类技能不需要注册,也不需要 API 密钥;只需安装 Python 3.8+,无需通过 pip 安装任何依赖。与 SEC 相关的工具只要求你在 trader/config.json 或 SECUSERAGENT 环境变量中填写姓名和邮箱。真正的下单则通过 Alpaca 的 MCP 服务器完成(默认为模拟账户)。
它能自己下单吗?不能。这套工具包中没有任何内容能自行发出订单:/trader-risk 必须传入确切的订单内容,你必须在聊天中输入 yes 确认,而且默认使用模拟交易。除非你明确开启,否则不支持期权、保证金、做空或杠杆。
Time Machine 技能具体做什么?它会使用该股票最近的 5 分钟K线数据,外加该交易设置过去每次出现时的全部交易日数据,把当天剩余的交易时段回放 10,000 次,从而在你入场前给出达到目标价或止损价的较为可靠的概率。
Pelosi Tracker 技能是做什么的?它会在 House Clerk 归档国会议员的股票交易披露文件的那一刻即时读取这些文件(根据《STOCK 法案》,须在 45 天内披露),让你看到的是公开可查的交易动向,而不是内幕交易。
这个工具真的能通过我的券商执行真实交易吗?可以,但只能通过 /trader-broker 完成。它通过 Alpaca 的 MCP 服务器连接你的券商(默认为模拟交易),并且仍需你确认具体订单,再在 Claude 弹出的确认框中点击 Allow once/Deny 进行二次确认。
为什么是十一个独立的技能,而不是一个大技能?每个技能负责日内交易中的一个环节(扫描、新闻、入场时机、出场规划、模拟、风险检查、仓位管理、下单执行、提醒、财报和国会披露),并把输出写入共享的 trader/ 文件夹,供下一个技能读取。这样整个流程既模块化,也便于追溯审查。
大小
0.4 MB
星标
8.4K
更新
提交Oct 20
时长
29天
提交 · 90 天